IUBio Biosequences .. Software .. Molbio soft .. Network News .. FTP

Logic of cladistics

Mark Siddall mes at zoo.toronto.edu
Sat Jun 11 07:41:04 EST 1994

In article <mezwick-090694153555 at pbmac-16.ucdavis.edu> mezwick at ucdavis.edu (Mike Zwick) writes:
>I have one question about the logic of cladistics.  Do you agree as other
>"pattern cladists" have suggested (rather strongly among some that I have
>had the opportunity to interact with), that their methodology is assumption
>and process theory free i.e. that they are looking solely at pattern
>(characters + parsimony and that is it)?  Would you agree with this
>It has always seemed to me that by employing parsimony, one is making a
>very specific assumption about the evolutionary process (assuming one
>believes that the evolutionary process is what is responsibible for the
>change of characters).
>If for example, in the analysis of molecular characters, one knew that the
>substitution process did not follow the process of parsimony, then would
>one be justified in employing parsimony to analyze molecular data sets?  I
>would think not - instead one would probably want to employ maximum
>likelihood (with a specific model of molecular evolution for example).    

This is a VERY convoluted issue and has a lot to do with history.  It
may also have more to do with professional loyalties than any 
substantive philisophical differences among cladists (see Hull's 
Science as a Process). 

Avoiding the issue of "pattern" or "transformed" cladistics to begin

No I do not believe that cladistics is assumtion free, nor entirely
objective.  Subjectivity come in at a number of levels including
choice of characters, interpretation of morphological character states,
sequence alignment (esp. when by-eye, but even passively when done
blindly by say Clustal).  
As for assumptions... the history of the whole busnessstands in relation
to the eclipse of phenetics by cladistics for the purposes of 
reconstructing phylogeny, and prior to that the eclipse of the 
synthetic or evolutionary taxonomy school by the pheneticists (BTW...
I do not use "phenetic" in a derogatory way).  The fetish for making
a big deal about assumptions started with the pheneticists objecting
quite rightly, to the ET-ists just coming up with a classification
on the basis of their good-judgement.  There was little if
anything empirical in it which led to highly plastic schemes and to
a situation where there was no repeatability.  This was not seen as
"science" (I agree).  Not everyone could play in the sandbox as the
"authority(ies)" could simply object until you'd acquired a reputation
as good as theirs, usually not possible unless you agreed with the
relevant authority in your field.
Pheneticists came up with character-based numerical approaches to the similar-
ities seen among organisms and devised (to their detriment, a plethora)
algorithms for grouping them.  Notably, Sneath and Sokal, in Ch. 8 of their
book Numerical Taxonomy indicated quite clearly that phentics was not
phylogenetic, and that it was not so-much geared towards the 
reconstruction of evolutionary history as it was towards the simple
grouping and classifying of taxa.
The cladists came along and said... hey!  We want phylogeny, and this
phenetic stuff we've been doing ignores the directionality of time
in evolution.  For phentic trees to be taken as phylogenies required that
there be ad hoc hypotheses about various character changes.

The cladists (like me) suggest that similarity is not primae facie 
evidence of RECENCY of common ancestry - witness crocodilians and birds.
But that the special similarities of descendant states, as opposed to 
ancestral states, are more important for determining phyletic

So on to assumptions...  I assume that evolution has occured, I assume
that it is directional (falsfied only if time is found to be non-linear...
and NO I do not want to get into a metaphysical debate).

What of processes....
Does adhering to parsimony mean you are assuming that evoltuionary 
processes are parsimonious?  I don't think so.  First, the resultant trees
will (almost) always have changes that are not parsimonious for 
particular characters.  Secondly, let's say that one cannot know the processes,
or does not wish to unecessarily invoke processes... the cladistic approach is
that the simplest explanation (i.e, parsimony) is the best, when given 
no reason to choose something more complicated.  Let's say you know a
bird left Albequerque and is now in Allamogordo.  You didn't see it fly
so you don't know the path it took.  If you draw a striaght line
you know that the distance flown at least isn't going to be any shorter.
But birds don't fly in straight lines... agreed, but of all the infinite
non-straight approaches, which do you choose?  You can't.  So hypothesize the 
straight line.  Or if there's a cactus in the beak of the bird and no 
cacti on the stright route, find the closest cacti to the straight
line and triagulate.  As Joe Felsenstein pointed out (not in this analogy)
if the bird is in Heart's Content Newfoundland, the straight line from
Albuquerque is more likely to be wrong than the shorter line to
Allomogordo.  Still, we agrue, you're not going to get a more efficient
So I don't think this is so much an assumption of parsimonious process
as it is an unwillingness to invoke a priori processes or explanations
that are untestable or unknowable.

My impression is that Joe's approach is one of: Well, if you're willing
to admit that parsimony is simply efficient, and is going to be wrong
on certain counts... why not try an approach that will tell how wrong.
Why not try an approach that invokes certain model(s) of process (like
building in rest, foraging, and geomagnetism into the bird migration
example) and discover phylogeny that way?

My concern, is that this then becomes a joint hypothesis.  Type I and 
Type II errors could result from data or from the model and you
cannot tell which.  That doesn't mean it's "bad"... after all, 
parametric stats and Monte Carlo simulations are all joint-hypothesis
oriented.  It becomes an issue of what the investigator thinks
is defensible about the system.

I don't know if that helps in terms of perspective.  I hope so.


Mark E. Siddall                "I don't mind a parasite...
mes at vims.edu                    I object to a cut-rate one" 
Virginia Inst. Marine Sci.                     - Rick
Gloucester Point, VA, 23062

More information about the Mol-evol mailing list

Send comments to us at biosci-help [At] net.bio.net